National Junk Food Day (202/365)

Posted on by

It’s National Junk Food Day!

As though the US needs an excuse to eat poorly.

Strange, I don’t think there is a Binge-Drinking Day—though many days in Wisconsin might qualify as honorary entries. Nor is there a Do-Controlled-Substances Day—though, admittedly, that one would be broad. I’m on prescription meds every day of the week. Sometimes when I write these blog posts, I wonder if they are really doing what they promise. And perhaps many of you, when reading my posts, also become paranoid. Apologies if so. Today’s post is a doozy…

As for junk food, my substances of choice are Kinnikinnick Vanilla-Frosted Gluten-Free Donuts. They come in a box of six, typically freezer-burned between the gluten-free lasagna that no one has bought in 11 years and the frozen tapioca bread that… well, no one has ever bought.

But these donuts—from Edmonton, Canada, nonetheless—sell fast, because they are as addictive as nicotine. My entire family concurs: these donuts are laced with something. Even completely frozen, I can gnaw through half a box in about two minutes and still be clawing at the back three like a rabid squirrel attacking acorns.

The downside? These things are kind of pricey. With inflation what it is, these small donuts come to about a buck-fifty each. And having grown up in the 1980s, when you could get a six-pack of Little Debbie Swiss Cake Rolls or Nutty Bars for 69 cents at the local convenience store, this seems outrageous to me. I sure as heck didn’t give up smoking in my early 20s to splurge $8 on donuts that last two minutes instead. By comparison, nicotine was an affordable habit… though I haven’t looked at the price of tobacco since 2005—in Hungary. Perhaps cigarettes cost more than $1.50 a pack now, too?

Alas… I guess what I’m getting at here is that I have had and continue to have vices. (I presume most people do.) I don’t drink. I don’t smoke. But I am an All-American junk-food eater. It is my weakness. My cultural shame. And I’m sure it will soon be my undoing, as my heart can’t take much more cheese, donuts, and ice cream. Nonetheless, if it’s true that people trade one addiction for another, I’m happy I landed on junk food. Because in America, eating processed food that’s bad for you is not only culturally acceptable, it’s celebrated with its own day! Hip-hip-hooray!!!

Wait… what is Halloween then, you ask? I can see a non-American’s confusion about this. However, that is definitely not an official junk-food day. That’s an annoy-every-person-on-your-block-and-consume-mass-quantities-of-candy day. Very different. That’s why we have both, I imagine.

Which raises the question: are there Fruit or Vegetable Days? I don’t know. I guess Ken and I will find out. Stay tuned.

It’s hard not to assume Junk Food Day was developed by some cabal of fast-food executives, soda companies, and chocolatiers—and I know, I know… I’m going to cut some of you off here. I’ve heard about those studies. “Dark chocolate is good for you,” some say, so I shouldn’t lump dark chocolatiers into this mix.

Whenever my kids tell me this bit about dark chocolate, I remind them about that infamous “Danish study” my Danish friend Klaus talked about nonstop during our senior year at Duluth East High School. He cited it like gospel. “Ian, a Danish study says…” And you can guess what a Danish study about beer consumption might have said in 1997: drinking three beers a day is really healthy!

Here is the funny thing, though. “Scientific research” says whatever the hell you want it to say if you stop following the research the moment you find a statistically significant result you want to hear. I didn’t know that then, so I couldn’t tell Klaus while huddled around a fire and drinking my third beer to my health. I get it now. Blasted… I later read an article disputing the Danish study completely. I should have stopped learning.

Which, of course, brings us to AI.

On LinkedIn, I keep seeing people post education studies showing that students who used AI in college last year are “dumber” than students who came before them. I’m paraphrasing, but not by much. The accompanying posts usually imply that humanity is losing its ability to think, civilization is nearing its intellectual terminus, and whither humanity if no one can read a passage of *Moby-Dick* and explain what drives Captain Ahab – or a fictitious whale.

The studies come fast and furiously out of education research centers right now because there is ample funding—and ample angst—around AI in education. AI is definitely going to flip traditional education on its head, or perhaps more likely, knock it on its ass. Some of the anxiety is warranted. Some students will absolutely use AI to avoid thinking. I avoided thinking with Cliff’s Notes, cooperative learning assignments, and by sitting near my smart friend Darren in grade school. In Hungary, my half the dorm spent three days creating cheat sheets and placing them in every conceivable place on their clothing, pens, etc. It was madness. They told me it was part of the academic process. Every generation develops its own technologies for intellectual evasion. Ours were simply less efficient and more analog.

But I keep seeing several different questions collapsed into one:

  • Does AI help someone produce a better answer?
  • Does using AI help that person learn?
  • Can the person reproduce the result without AI?
  • And is reproducing the result without AI still the ability that matters?

Those are not the same question. Not even close.

A student can produce better work with AI and learn very little. That is possible. In fact, a large high-school mathematics experiment found that students using unrestricted GPT assistance performed better during practice but could perform worse once the assistance disappeared. (Shocker!) Yet, a more carefully designed AI tutor reduced that problem. In other words, “AI harms learning” and “AI supports learning” can both be true, depending on what the tool does and how the student uses it. How terribly inconvenient for LinkedIn.

There is also evidence that generative AI can improve higher-order problem solving, and workplace studies have found substantial productivity gains—especially among less-experienced workers. That fascinates me. AI may not magically pour wisdom into everyone’s skull, but it can make some forms of hard-won expertise available to people who do not yet possess it. That is not the same thing as becoming educated. It is not the same thing as becoming intelligent, either. But it is not nothing. It may turn out to be economically enormous.

Education studies evaluate educational achievement. Intelligence research studies cognitive abilities. Workplace research evaluates performance and productivity. These things overlap, obviously. Education can strengthen cognitive abilities, and intelligence strongly influences educational achievement.

But educational achievement is not synonymous with intelligence. A decline in math or reading scores does not, by itself, prove that an entire generation has become less capable of reasoning, adapting, inventing, navigating social life, or solving problems with the technologies available to it. It proves that performance declined on the abilities being tested. That matters, but it is a narrower claim than the cultural obituary people like to write from it.

The more interesting question is whether we are measuring the abilities people will need next, or perhaps more aptly, the abilities institutions needed people to possess yesterday.

Shocker: replacing schools with hastily assembled Zoom links, parental panic, social isolation, and two years of Fortnite did not produce a pedagogical renaissance. Math and reading scores went down. But I remain curious: if we had also measured eye-hand coordination, rapid visual processing, group strategy in distributed environments, and the ability to pilot a drone with an Xbox controller, might some scores have gone up? I’m guessing yes. Why do we assume the skills schools already measure constitute the natural borders of human intelligence? With the rise of AI, they are far less likely to measure the skills today’s employers will be looking for.

As a daily mapmaker of products like the National Retainer Day Map, I know better than to mistaken the borders on a map for the borders of the world – particularly gnomonic-projected ones.

This is where the philosophy gets interesting – or perhaps unbearable – depending on your tolerance for philosophy in a post that began with frozen donuts.

Humans have always placed parts of our intelligence outside our bodies. Writing stores memory. Mathematics stores relationships. Libraries store collective knowledge. Maps store spatial reasoning. A map allows someone who has never stood on a mountain to understand the valleys around it, at least partially. No one looks at a map and says, “Well, this is disturbing. Humans will lose their innate ability to survey a continent personally.” The map is understood as an extension of cognition. It’s not proof that human spatial cognition has ended.

Socrates reportedly worried that writing would weaken memory and give people the appearance of wisdom without its substance. The annoying part is that he was right. Writing did weaken the need to memorize. (That’s why we aren’t reciting The Odyssey to one another but going to watch it in cinema.) But writing made philosophy, science, literature, law, and this unnecessarily long blog post possible. Technologies very rarely make us simply smarter or dumber. They redistribute our cognition. We stop doing some things, begin doing others, and then pretend the new current arrangement has been natural all along.

Every major technology of thought has allowed humans to forget something; it’s wonderful! We use the freed cognition to think about something else.

There is an old philosophical idea wrapped in the Greek word “pharmakon”: something can be both a remedy and a poison. The dose, context, and relationship matter. Junk food is not poison in the same way arsenic is poison. A donut does not destroy nutrition. A diet consisting entirely of frozen donuts, however delicious and economically ruinous, is another matter.

AI is probably a “pharmakon” too. It can be cognitive junk food: frictionless answers, consumed rapidly, providing the sensation of understanding without much intellectual nutrition. It can also be a kitchen, a tutor, a laboratory, a sparring partner, a drafting table, a coding collaborator, or access to expertise that a person previously could not afford. The fact that it can be abused tells us almost nothing about whether it should exist. Most useful technologies can be abused. Have you seen automobiles? Social media? Your fellow humans?

The question is not whether AI makes humanity smarter or dumber. The question is which forms of intelligence it amplifies, which abilities it allows us to neglect, and which neglected abilities will turn out to have mattered.

That last part is important. I am not arguing that knowledge is obsolete. You need enough understanding to recognize when a machine is feeding you nonsense. You need judgment. You need context. You need skepticism, curiosity, and some idea of what a good answer looks like to you. If anything, AI makes those abilities more important because it can produce incredibly incorrect answers that sound… well, educated.

Personally, I think formal education will lose its monopoly over expertise – at least partially. For generations, education served at least three roles: it helped people learn, certified that they had learned, and rationed access to professional opportunity (i.e., career). AI disrupts all three, but especially the rationing. When expertise can be called upon in seconds, knowing everything personally may become less valuable than knowing what to ask, what to distrust, what to combine, and what to do next.

That does not mean math no longer matters. It means the economic value of manually performing certain mathematical procedures declines. Calculators did not end mathematics; they changed which operations deserved human attention. Spreadsheets did not end accounting. GIS did not eliminate geography—though it did tick off a bunch of human geographers for quite a while. AI will likely do something similar at a much broader scale.

Will my daughters need calculus if they become engineers or work on global map projections like my friend Bojan? Probably. Will they need to execute every derivative by hand for months to prove they understand the concept? No. The durable ability may shift from successfully executing procedures to framing problems, selecting methods, inspecting results, and noticing when the answer violates reality. Iteration… collaborating with intelligent systems. That will be important.

Writing will change in much the same way. I find it hard to believe that most consequential final drafts will not involve AI within five years. I may have written my last book using the archaic method of organizing, drafting, and writing every word one after another myself. Am I sad about that? A little, perhaps – kind of like I miss making proportional symbols by hand in Macromedia FreeHand. I look back on it fondly. I enjoyed the craft. But I try not to confuse my affection for past techniques as proof that everyone must preserve the workflow forever.

Imagine if I had spoken this free-write into ChatGPT and asked it to organize my thoughts, check the claims, streamline the digressions, and make sure I didn’t offend anyone. You would probably be appreciative right now, and what I’m rambling about might make more sense. We would both win: I wouldn’t get cancelled, and you would be able to understand me. What’s not to love?!

Well… something might be lost. The rhythms might become too smooth, right? The jerkiness of driving a Chevy Sonic stick shift giving way to a VW Tiguan automatic. The weird connections might disappear. The machine could politely remove the frozen donut and Socrates references I added here, because they do not appear relevant to my – or its or our? – thesis. The resultant write-up may be clearer for you but worse in affect. In writing, efficiency is not the only value, and polish is not the same thing as meaning.

That is why human direction still matters.

What about coding? Many software engineers believe coding is indispensable, which is understandable because it pays their mortgages. But the history of higher-level programming occupies a mere blip in human history. It exists largely to let humans communicate instructions to computers without arranging ones and zeros personally, which would be a terrible way to spend a Thursday.

Now we can increasingly communicate with computers in natural language. Why will everyone continue choosing a syntax-strict artificial language when ordinary language can express the desired outcome? Perhaps coding, as most people currently practice it, was a transitional stage in our relationship with computers—the first date, as it were. We have now entered the deep-talking phase.

That does not mean software engineering disappears next Tuesday. Natural language is ambiguous. Machines misunderstand us. Systems still require architecture, testing, security, verification, maintenance, and people who understand the machinery beneath the charming conversational surface. But the balance is shifting from manually constructing every instruction toward directing, evaluating, and integrating machine-generated work. I see no persuasive reason to pretend otherwise because the old arrangement made expertise scarce and salaries high.

Some people would say the future I’m describing is dystopian and ridiculous. Maybe. Four years ago, I would have said what I’m saying now was pure bananas. But I no longer believe refusing the technology will preserve human intelligence. Humans are technological animals. Our intelligence does not exist separately from language, tools, symbols, institutions, and one another.

The serious risks are not that AI is unnatural. The serious risks are that a handful of companies may control our cognitive infrastructure; that systems reproduce bias and error at enormous scale; that people surrender judgment for convenience; that surveillance becomes ambient; and that productivity gains flow upward while displaced workers are told to learn prompt engineering. Those are political, economic, and ethical problems. “Ban the chatbot and make everyone write five-paragraph essays again” is not an adequate response to them.

Perhaps the reason many of us – including my entire family, except me – fear this future is that we grew up when humans had to accumulate education to become the experts who built technology. Expertise lived in people and institutions, and access was scarce. Now more of it can be summoned on command. That does not mean the machine contains all human intelligence or possesses wisdom. It means the practical boundary between “what I know” and “what I can competently accomplish with assistance” is becoming harder to locate.

And perhaps that boundary was never as firm as we imagined. How much of my intelligence resides in my brain, and how much resides in books, colleagues, software, databases, maps, search engines, and the accumulated work of people I will never meet? If you remove all external assistance, are you measuring my pure intelligence – or merely placing me in an artificial and rather pointless survival game?

That is why I encourage my daughters to use AI regularly. Not to avoid learning, but to learn how to work with it while they still have teachers and parents around to tell them when it is bullshitting. They are already smart, and they are doing well in school. Education will continue to matter, although perhaps less as a gatekeeping credential and more as a foundation for judgment. They may enter an economy that rewards them not only for what they know, but for the quality of the questions they ask, the problems they can frame, the tools they can direct, and the compassion they show toward people affected by their decisions.

I hope so, anyway. Their parents have made a living largely from education and expertise, and we may be quite worthless in the market by then if I’m right and the human-expertise job market starts downsizing. Every utopian forecast becomes more interesting when the author’s own paycheck is fed into it.

What a trip… from Junk Food Day to Danish beer studies, Socrates, the externalization of human cognition, and my optimistic – though some would be forgiven for thinking dystopian – thoughts about life with AI.

This whole essay is speculation written to intentionally disturb coming to any comfortable conclusions, including my own. I go where the evidence leads; I hope others do too. And I”m often wrong; I chose Microsoft Zune over iPod, for example.

Should writing everything manually in a word-processing app and high demand for coding boot camps come roaring back, I’ll concede that I was way off. I would even bet against it right now, but gambling is not one of my vices. I have to draw the line somewhere.

Did I digress…?

A daily map. Which one is it again? Oh yes! Junk food, naturally.

For today’s map, I created a bivariate map of the two places where I most often access my junk food: convenience stores and fast-food restaurants.

I also made it using a Macromedia Flash MX map-design schema. And, strangely enough, the map demonstrates the entire argument I have just dragged you through. (It’s weird that if you write long enough, and it hits 1:30 in the morning as it has here, your argument may just naturally come full-circle.)

Notice – which will be particularly hard to miss if you are red–green colorblind, I imagine – that the map uses red and green quite inappropriately. Not good. Horrible. And I realize this fact.

But I left it because the schema was accurately reproducing what many Flash maps did in the early 2000s. ColorBrewer had not yet appeared, or had just appeared, when I made my first Flash maps. Red and green were still frequently placed together. Red meant stop and green meant go, after all. It seemed intuitive to many designers at the turn of the millennium.

The AI followed the design schema I fed it. It reproduced the period accurately. And the result contains a poor accessibility decision.

That is the point!

Successful execution is not the same thing as a wise outcome. A system can follow its instructions perfectly and still produce something we should question. Prompt cartography supplies capability. The human mapmaker supplies – or ought to supply – judgment, context, responsibility, and the willingness to say, “Yes, that is historically accurate, but perhaps we should not do it.” (And of course, when someone tell me perhaps we shouldn’t do it, even myself, I tend to say: well, let’s see what happens if we do do it. I guess I can’t get enough null-hypothesis testing in my life.)

We have come a long way in twenty-some years – and over the 20 minutes you’ve been reading, perhaps). We will presumably look back at today’s AI products with the same combination of nostalgia, embarrassment, and disbelief. I hope we do. It will mean we kept learning.

Please use AI thoughtfully and eat junk food responsibly. There are children about!

Tomorrow is Rat Catcher Day.

God help us all. You can’t make these days up!

A Studies Fueling My Ramble

View Map Here

 

Web Mapper Prompt (Link)

Attached files: Flash_2000_v2.json, census_dataset.csv (created with Spatial Data Doctor)

HI. Tuesday is National Junk Food Day in the USA! Woohoo!

I would love to create an interactive web map to celebrate this triumphant day.

I’m attaching a dataset showing how many fast food restaurants there are per 1000 residents by US county, as well as how many convenience stores per square mile or similar, I forget.

Can you create a multivariate map that shows both variables at the same time, please? When users hover over a county, or click, please have an info window show certain relevant statistics you find in the dataset about the county and state that it is in overall. Give the map a title that makes it sound serious: Junk Food Access by County, etc.

Exclude superfluous decision and methodology text from anywhere in the map except for a “?” Help and Attribution modal.

Oh, finally, I just created a super detailed new map design schema that I’d like to try out. Can you follow this schema to design the map overall. Please note that this should work on both desktop and smartphone devices, but how it is designed for both may vary significantly.

Thank you!